I just found a workflow that makes AI subscription expensive so it doesn't make sense 😭
This guy stopped paying Claude $200/month and switched from Opus 5 to DeepSeek V4 Flash which he said was only about 1/20 the cost. There's only one problem: the model is blind. 👀❌ Can't see
DeepSeek Harness is finally here.
It feels so different that claude code & Codex, and the quality per token it drives is insane. Pasted a picture of traceable agent queries; you can literally look at everything generated and pinpoint the problems in your code/instructions.
DeepSeek Harness’ latest evil fix! Is the minimalist model that has been touted as a nuclear bomb really viable?
🤔
This is the best trick to maximum usage limits on chatgpt codex
Codex's best kept secret is that your main agent doesn't have to do everything... custom agents are just files in ~/.codex/agents, and one file gives you a second worker on deepseek v4 flash create
You have reached the end of the archive
All of deepseek